Transcript of the
Friday Discussion, January 18, 2008 at the CUNY Phonology Forum
here for an audio file of the Friday Discussion.
NOTE: There are a few points in here where a word or phrase
was hard to hear, and where speakers were not identified. If anybody can fill
in the missing words or names, please send an email to syllable at cunyphonologyforum
Chuck Cairns: Okay, so shall we commence with ...[today's
discussion]? And those are the
questions that we used for the call for papers, and that's just sort of to
prime the pump, so I'll just call on people and make my own comments, so here
we go. Oh, and try to remember to say who you are, because we are recording
this and we want to put it all up on the web. I hope you don't find that
intimidating or anything like that. Harry? I've already identified you, okay?
Harry van der Hulst: So, do syllables exist? It seems to me,
if I can go back to what I was talking about, I was mostly emphasizing the need
for onsets and rimes as different branching constituents. And something I said,
we are clear here on the issue of whether there should be constituents and
embracing those two, and so... And then I said, well, some of the phonotactic
facts seem to be best accounted for by assuming these types of constituents.
And then there are other phonotactic facts that are accounted for in terms of
these structural relationships, that was the second part of the paper, which
are linear statements. So we seem to agree that what... and I will agree with
anybody who will say that you need, let's say, linear type statements to
account for phonotactic facts, but then I maintain that that account doesn't
cover the whole story, and that these traditional, apparently Andrew said I was
going to be controversial, but I wasn't controversial about being very
traditional then in that case, but in the conext of this conference, it seems
controversial to say that we do need those branching constituents as the basis
for sorting out the phontactic facts that are there. So, perhaps indeed nobody
has argued strictly for the existence of the syllable as a constituent, but
some have argued for these onset and rime units that we might just as well call
phonotactic constituents, so I would maintain that these are real.
Chuck Cairns: Paul? Paul Kiparsky?
Paul Kiparsky: Could I just make a suggestion about the way
in which a discussion like this...
Chuck Cairns: Yes. Good, thank you.
Paul Kiparsky: ...a future discussion like this, might
proceed? So if we take, let's say, syntax, and somebody proposes a new
syntactic theory, there are certain basic phenomena that are agreed upon that
they have to deal with: raising, control, and so forth. And you don't have a
theory of syntax unless you handle that. So, I'm wondering if we could agree on
some core phenomena or generalizations that would be important to address, that
have been at least on some syllabic theories understood to be explained by syllable
structure. So anyone who proposes a theory of the syllable or a non-theory;
rather, a theory of the non-syllable, let's say, should provide an account of
those things, so those might include, for example, the well-established
typological generalizations, maybe the basic syllable typology. It might
include the generalizations about phonotactics about, perhaps, compensatory
lengthening, and maybe some core facts about English or other well-known
languages, let's say, Kahn's stuff. And the people who introduce new ideas
should take care to have a story about those, otherwise we end up talking past
each other, and I have a sense that in some of the talks, that was happening.
Chuck Cairns: Can we do that here now? I mean, can we...
Shanti Ulfsbjoerninn: I feel slightly guilty of Paul's
comments in a way, and I understand the need for having a kind of unified core
of facts, which would say, at least for dealing with phonology and something
else. But I will say that that can't be done by taking for example like, let's
say, the theory must account for, let's say, branching codas, when not every
theory of phonology accepts that these even exist, so I think that one would
have to be very careful about the kind of criteria that one has to meet,
although I do agree with a kind of syllable, basic syllable typology like the
example I've given. If your theory actually excludes CV segments, then you're
in trouble. So I just promote caution in choosing what things to use as
validating your theory.
Anthony Lewis: A starting point might be just something very
basic, like: why do we seem to see lenition and consonant deletion in
syllable-final position and not in syllable-onset position? As a starting
point, in many languages.
Eric Raimy: May I make a friendly revision to that? Why do
we find lenition, or the phenomena that you asked in particular contexts? So
remove the discussion of onset versus codas, because that...
Chuck Cairns: It loads it.
Eric Raimy: ... the facts of the matter is that, what we see
is that lenition occurs in particular contexts and not in others. Some people
would say, "Ah, that's a coda context. That's why it happens there. In onset,
it does not." But that, as Shanti pointed out, that biases the discussion
immediately by building syllables...
Chuck Cairns: Yeah. Donca?
Donca Steriade: So just continuing Paul's remark, I think
that the critical fact we need to address all is the possibility of convergence
between that the various phenomena that have been explained in terms of the
syllable are telling us about where the syllable is, is where it begins, where
it ends, and what internal structure it has. So I think that a lot of the
literature has pursued that the hope of explanation through the syllable as the
tool of the explanation, without checking that we're getting convergence on the
same boundries for the syllable that the facts about rhythmic alternations
between light and heavy syllables need to converge on the same syllable cut
that the phonotactic explanations require. And frequently we're not getting
that convergence, so it seems to be that, yes, we do need to understand what
any new theory's account of the old phenomena ought to be, but we also have to
make sure that these phenomena are converging on telling us that the syllable
is in the same place, in order to explain each and every one of them.
Eric Raimy: Actually, can I make a friendly revision to that
Chuck Cairns: Yeah, yeah.
Eric Raimy: So, I think we need to ask the question of what
parallels of convergence and non-convergence on evidence can we find in other
parts of language, because if we look at lexical access, it appears that
language looks very discrete. Digital, classical phonemes appear to be the way
that it works, because apparently that's what lexical access does. But if we go
look at socio-phonetics, now it looks completely gradient, because not only do
we have to have information, very subtle phonetic information about different
judgments about social class and gender, but we have to nail it down for
individual speakers. So it's not that there's a problem with the syllable in
the conflict, that we're not finding convergence of the results. It's the fact
that we don't have any idea how to get convergence on the results of language
because we always try to meet it as one model, one thing. So we have to kind of
parse out, just kind of like with Marie-Helene and her student giving the
posters, we have to ask like, what is the exact task? What is our exact probe
probing for? And recognize that if we use a different probe, we may not be
probing the exact same thing. So we shouldn't necessarily expect every
different type of probe we have to give us the same answer. We have to have a
better understanding of what the probes are doing and why we get different
conflicting answers. It's a much harder task, I admit that. But that's the only
rational way that I can see it goes, because you're right. I don't see
convergence in the field because you read one phenomenon from one viewpoint
using one methodology, you get A. The same phenomenon, different methodology,
different view, you get not A. And it's the same phenomenon, so it indicates
that both sides are wrong, and there's something in the middle that's right
about it, not that everybody's right or everybody's wrong.
Donca Steriade: Well, you see, it seems to me that the great
hope that we are all inspired by when work on the syllable began in earnest was
precisely the hope of convergence of multiple phenomena in one representational
unit. Insofar as that hope has sometimes been dashed, (interjection:
Sometimes?) we need to understand why that is. So if that was...
Eric Raimy: Okay.
Chuck Cairns: Harry?
Harry van der Hulst: So, despite what was just said, and
also, I agree with what Paul says, that we should, we can perhaps ??; suppose
we had never heard of the syllable, that all of us, and that we have been in
this conference listening to all these phenomena, people discussing them,
having explanation so, and explanation this and that, and somebody would come
in and say, "Look, I think that we can explain a lot of this stuff by this
particular thing, this notion. Perhaps not all of it." I guess we all would
have applauded him and given the person a Nobel Prize for phonology. You know,
you are on the right track here. It does unify a lot of the stuff. So, I mean,
that's my feeling and that's why, I guess I want to stop arguing and clinging
to that notion, even though I agree that there are other quality explanations
that have... it doesn't account for everything... And it's always good to question
your assumptions and so on and so forth, but that's still the feeling I've got,
that you can desperately try to avoid it and pull in this and pull in that, and
do this and do that, but I still get that feeling that, Donca said, that old
feeling that things started out with, that there is this thing that brings
things together. But okay, I mean, is that true? I could speak about a couple
phenomena in Dutch, a well-studied language, like English, regarding
phonotactics and stress, where the particular assumption that, you know, that
/br/ clusters as a unit, creating an open syllable to the left explains why the
vowel to its left has to be tense. It has to be a tense vowel. The word for 'zebra'
in Dutch, /zebra/ has to be like that: the first vowel has to be tense. That
same assumption accounts for the fact that, when you have /br/, that the
preceding syllable in fact is open and therefore light for stress. So these
things totally converge. And can you avoid it? Yeah, you can avoid it but then
you get the same problem that Venneman was talking about into the 1970s that
you repeat certain observations both in your statements about tense and lax and
in your statements about stress, which was the initial, you know, force of
arguing that perhaps there is something there to begin with, that both of these
phenomena make reference to. I think that there is such convergence in
languages which, I guess, is the reason why I may want to be controversial by
Chuck Cairns: Bill?
Bill Idsardi: Okay. I guess my feeling about all of that is
sort of to be methodologically anarchistic with respect to that. So, I think
the problem is, I agree with Paul that you want to eventually account for all
the things that previous theories have accounted for, but, you know, there are
cases where that sort of requirement just leads to an inherent conservatism. So
like, the, in the speech recognition literature, you know, you have this thing
where the hidden Markov models have been so prevalent because you have got to do
one hundred percent better every year. And so it's been very difficult to sort
of, break that out of the literature in that case, kind of because of the way
that goes, and if you're not doing as well as the other groups, you know, then
you're just nowhere. I mean, so you can't play the game unless you ante up the
whole way. And I think the other things are true too, so as Harry said, it's
like you've got this convergent set of things, which in SPE would have been
handled by talking about the weak clusters or the strong clusters, without
reference to any kind of syllable structure, and then the question is: How do
you evaluate those things? So if you have a pattern of generalizations, does
that mean that you have to have a representation? Or do you have something like
SPE used, which was a series of markedness statements and simplicity metrics
that allow you to say, "oh, this grammar is more highly valued than that
grammar," by reusing statements in various ways. So like, you know, there was
essentially a procedure that looks like something like minimum description
length, where you're able to say, "okay, I've seen this string a lot. I'm
allowed to call it A and then I'm allowed to make generalizations about A." But
that doesn't imply necessarily that there is a structure associated with A,
beyond just that string thing. So, as far as I'm concerned, there's no way to
know going in whether any of those argues directly for a representation. You
have your account and your account works or it doesn't work so whatever... and as
you pile together your things, you group them in different ways and other
people group them, well, you're going to have convergence on one thing and
divergence on something else. I don't see any particular way... I don't, I think
it's a general problem in science. I don't see any way of knowing going in that
this convergence is better than that convergence. That getting A and B grouped
together is better than getting A and C grouped together.
Chuck Cairns: Paul?
Paul Kiparsky: Yeah, I just want to add an aside to that.
The only case where the minimum description length in phonology was
consistently and precisely applied was Panini's grammar, where literally every
technical, every concept, every category had to pay for itself. The definition,
which was the part of the grammar, had to cost you less than the amount of
savings that the use for that definition across the system simplified the
description. And it turns out that all the interesting theoretical notions that
have emerged from that: theta roles, and you name it, it's there. Ordered
rules, all of the wonderful things, except, the syllable didn't. He managed, he
managed without the syllable. (Inaudible interjection)
Darya Kavitskaya: Okay, the question which I have is, we are
trying to answer the question, do syllables exist? But in what sense are we
trying to answer it? I mean, we are trying to answer it in many senses. We are
talking about syllables being in existence from the point of view of phonetics,
but we don't see any evidence in articulation. Well, maybe we do because there
is some, there is a collection of emergent evidence in articulatory phonology
which sort of points to the existence of some kind of unit. We don't see
evidence in acoustics. We are trying to debunk all evidence we see in phonology
but, as Harry says, sometimes it's a bit hard to debunk even though we can show
that it's not there. That with our native speakers' intuitions, and native
speakers' intuitions about syllables are there. What's not there is native
speakers' intuitions about syllable boundaries. I mean, this is variable. So,
do we treat this kind of evidence as being the evidence for the syllable? Or do
we say, "okay, fine. Native speakers of my language and many other languages
will tell you what the syllables are in their language. They'll calculate
[claps?] Maybe they will say the vowels. Sometimes they'll flap their
consonants. But in many cases ...[?] So that's the question. What are we looking
for? In what sense do syllables exist?
Anthony Lewis: Just a follow up. Can't we usually predict
when native speakers will have problems in dividing syllables to a certain...?
(Darya: Yeah.) Then, what does that tell us about syllables that we can then
predict... let's take cases in English. So nobody has a problem with 'hotel'. At
least, that's what I would assume. But there are other cases like, you know,
'rabbit', where you'll get all sorts of answers. So we could kind of, so what
does it mean that we're able to... and probably in Russian or Polish that has
these clusters you probably can predict, where there are situations where
people want to find evidence and end up basically agreeing where there's
there's going to be variation.
Darya Kavitskaya: There is, if I may, in Georgian there is
also a situation with huge clusters, but in Georgian, people usually don't have
any problem dividing things into syllables, except for sometimes, but it's not
like in Russian.
Chuck Cairns: Harry?
Harry van der Hulst: Bill's point about referring to
recurrent sequences without calling them constituents, so would you, Bill, then
extend that argument in the domain of syntax? So would you then stop teaching
people about constituents there and say, well no...
Bill Idsardi: No, I didn't suggest stop teaching people
about them. I mean, I just... what I was arguing for was the anarchic position
that you have to consider both positions, right? And there are certainly people
out there, I mean not in linguistic departments, but, who would extend it to
syntax. I wouldn't be one of them, but I think that it's not a bad practice to
go back to first principles and say, if you've got this other explanation
available to you, how much do you have left over that your theory gets that
other peoples' don't? I mean, I think one of the cases that drove this home for
me a long time ago was Mitch Markus's sort of discussion about what Roger Shank
had been doing. And you know, his conclusion was, it was amazing how far you
could get with Shank's theory. I mean, there were some things you couldn't get,
but it was amazing how well it actually did, I mean, given how terrible it
Harry van der Hulst: And that's another thing about... certain
people find it instructive to learn that boundaries between syllables are
difficult to detect or that we have no intuitions about it. Now, I find it
difficult to detect exactly where my nose starts and where it ends, you know,
looking at my face, or where my ear starts and where it ends. But that doesn't
take away the fact that, I suppose, in a biological sense, these are separate
organs and perhaps even, I mean, at least I'm not a geneticist, but maybe at a
genetic level, maybe there is a separate set of instructions there for building
ears and noses, but it works out that the edges blur in the actual phenotype.
So that fact that, if you think of phonetic syllables the way that you can
measure them and the way that people can reflect on them, are they reflecting
on phenotypic things and we have no intuitions about the structure of a
nucleus, so to speak. So I'm not impressed by the fact that these boundaries
are not easily detectable, because, you know, that happens in a lot... by that
same token, we should all abandon segments right away, because there's no
boundaries there, at least not in the sense... you know, that's what we always
teach our... you know, one stream of sound. Maybe if you are a trained
phonetician you can see things in a spectrogram, but, so we should abandon all
these distinctions right away, because we have no intuitions about where these
things stop and begin or begin and stop.
Eric Raimy: Actually, I was going to say, I think a point of
unnoticed convergence in here is that, the issue that I think we're having
discussion here is that, we all agree that there is some sort of distance of
something in between recognizing points in the speech signal. And, the syllable
approach basically says, "I have to be able to divide that in some way and then
have stories of how each half of it goes." And then the minimal string length
approach says, "Well, I don't care what the span is between these two points
are, if I see it enough or if I can just randomly encode it." And so, I think
we do share some consensus that, I think, if we model kind of like what we know
about stress, is that counting one thing's okay, counting two things is okay,
counting nothing, so unlimited like the Bella Coola is okay, but if we begin to
have theories saying you're not allowed to have three or four or seven things
between the regions, that's an unlikely syllable length, because it seems to me
that all of the transitional stuff is very local. You know, it's always going
into a sonorant or coming after something. It's all very local and so counting
gets limited, all right? But, we know there's a limit to these long-distance
effects. So, Stephanie was talking about the Abercrombian foot, which appears
to put three syllables in a metric constituent, but we're counting off of
stressed syllables to do that, so once again it's a local one. So I think one
possible place of convergence is actually seeing what possible kind of
calculations we feel comfortable with allowing and if the syllable approach
accounts for it one way as it kind of counts in from the edges, and the string
approach will define legitimate strings from other things. And so, I think this
is, maybe I'm reading into people's interpretations of what they're saying, but
this seems to be a common theme that everybody shares, is that nobody wants a
constituent, however we define it, of five arbitrary things. But one arbitrary
thing, two arbitrary things, or you know, no constituent, or anything appears
to be fine.
Chuck Cairns: I would like to interject. I would like to ask
Paul if he would accept this paraphrase of sort of the challenge you made to
anybody who would propose the non-existence of syllables or at least extremely
barely minimal syllables. You listed a bunch of phenomena, like for example,
various kinds of typological stuff, all of the stuff that Kahn talked about,
like the distinction between you know, in 'mattress' versus 'atlas', the
aspiration/glottalization phonotactic stuff and stress attraction and so on. I
think everyone has to agree that all of those things are within the explananda
of phonological theory. You have to have a story for those. It doesn't
necessarily follow that the syllable has to be the story for those. So anybody
who proposes a phonological theory that does not include the syllable or has an
extremely barely stripped-down syllable has to have at least some kind of
plausible proposal that, all of those other phenomena that we had taken under
the Kahnian and its various kinds of descendent theories of the syllable, we
have to have a way of approaching those. We have to have a good story for
those. So, we, you know, we don't want to lose things out from under our explananda
as we propose new ideas. But it's sort of like, I hope this doesn't offend
anybody, but it's sort of like the argument that atheists frequently put
forward, that theists have argued for centuries that we need the notion of God
to explain all of these various phenomena that we see out there in the world.
Well scientists come along and sort of carved away these various things and
they turned out not to be into a natural class and the number of phenomena that
are left over that you have to be able to explain by invoking God gets smaller
and smaller and smaller. Maybe the syllable turns out to be sort of like that.
Paul Kiparsky: Well, one way to look at it is that the
description for a _____ theory has to support that, so one of the problems with
Optimality Theory is that as the theory gets deeper and deeper, there's fewer
and fewer ways of actually writing a grammar, so you haven't seen this great
flood of Optimality Theoretic phonological descriptions of entire languages.
Chuck Cairns: That's right. Exactly, exactly.
Paul Kiparsky: It's usually just a couple forms, so it's
very deep, but there's no coverage. What I'm worried about is that all these
proposals for the non-existence of the syllable would have some of the same
effects, that is: how does an actual phonological description look like based
on that? I want there to be lots and lots of grammars of languages, including
bit fat parts about phonology and any theory that allows that to go on is fine
by me at some level. I also want the deepening, but I don't want that to come
at the price of just ignoring everything.
Bill Idsardi: I think that's an excellent example of this
tension between typological coverage versus depth of coverage in individual
cases. You want them both but I don't think there's anything to see that one
way is the way to go and the other way is not. You just have to do them both.
Chuck Cairns: Stephanie.
Stephanie Shattuck-Hufnagel: It strikes me that if we really
wanted to do what you just described, and what I think Paul is also proposing,
it might be extremely useful to have a list.
Chuck Cairns: Yes. You mean a list of things that we want to
Stephanie Shattuck-Hufnagel: What has the syllable been
doing for us all these years that we don't want not to be able to do if we give
Chuck Cairns: Yes.
Stephanie Shattuck-Hufnagel: Maybe there could be a site
somewhere that we could park things there.
Chuck Cairns: Well, there's one. Oh, that's not one. CUNY
Phonology Forum is one. I would love to promote that discussion. Darya.
Darya Kavitskaya: If we had that list, and the explanations
from your non-syllable theory, that would explain all these things differently, have we achieved anything?
Stephanie Shattuck-Hufnagel: I think so.
Darya Kavitskaya: What have we achieved?
Stephanie Shattuck-Hufnagel: What we've achieved is the
knowledge that there's more than one way to account for the things we know so
far, and therefore we need more things to account for in order to distinguish
which of these approaches is more appropriate.
Darya Kavitskaya: So, at that point we know that we don't
Harry van der Hulst: Well, you achieve something if all
those other explanations are independently motivated.
Chuck Cairns: Yes. That's right, exactly.
Harry van der Hulst: Then we can get rid of God, the
Donca Steriade: I actually don't understand that. I mean,
quite frequently when you... so the initial explanations in terms of the syllable
for a phenomenon, they use the syllable as a descriptor, as a compact
descriptor of the context in which something happens. So, in German, we get
devoicing at the end of the word for obstruents. So now, it's syllable-final
devoicing. It's a shorter statement than that. Do I understand why in
syllable-final position I have to devoice and not in syllable-initial position?
No, I don't understand that. So I think that on that website that Stephanie's
proposing, we ought to... clearly it's important to have compact descriptors, but
we also have to understand that when the syllable is [cough] to us, why things
have to be that way and why they can't be another way.
Harry van der Hulst: Or maybe the mind isn't interested in
explaining things and wants these short cuts and wants these descriptors?
Donca Steriade: Well, we don't know.
Harry van der Hulst: Well, we don't know. I know we don't
know. I mean that's...
Donca Steriade: We don't know whether learners are looking
simply for compact descriptors, or maybe they are also looking for quality
explanations. (Interjections) We ought to keep an open mind about that.
Chuck Cairns: I want to make a procedural proposal. What I
would like to do, and I want you people to guide me on this now: in the next
few days, maybe a week or so after this conference, I will go through some
basic literature, like Kahn, and then maybe the chapters in the Kenstowicz
textbook, maybe couple of other basic places like that, and make a list of what
appear to be the basic phenomena that the syllable has been put forth to
explain. And I will make a link when you go to the CUNY Phonology Forum
website. I know you all probably have that bookmarked, I hope. Where I have
"Syllable Conference," you will click on that and I will have that out there as
a list, and then maybe we can take it from there. We can have an online
discussion that we will continue now, but... what do you people think of that
San Duanmu: Sounds great. Are you going to include the
distribution patterns of entire lexicons?
Chuck Cairns: Now that you mention that, of course.
Eric Raimy: As a procedural matter, I think how it should be
seen is that, once this is over, this is being tape-recorded, so we'll try to
transcribe what's occurred and pull anything that's said in this room as part
of the starting point. And then, it's up to everyone here that if we couldn't
hear or whatever, or if the transcript doesn't have the suggestion or the thing
that you think needs to be covered, you need to get involved: check the website
and add to it, and we'll figure out ways to make it as easy to kind of
Chuck Cairns: Yeah. Shanti.
Shanti Ulfsbjoerninn: Can I find out whether people think
that this is a paradox that we're kind of struggling with? It seems to me that
whenever you have any formal constituent, the reason, the fact that we call it
a constituent is because certain phonological processes identify it in
opposition to any other. So essentially, when we ask what phonological
processes show us the syllable exists, and why is it the syllable and not
anything else, it's kind of paradoxical, because the answer is that the
syllable exists because the phonological processes identify it. Do you see what
I'm getting at?
Chuck Cairns: Yeah.
Shanti Ulfsbjoerninn: So how do we actually find out what
kind of evidence we're actually going to accept that it's actually a syllable
and not just anything else like an individual context?
Chuck Cairns: Exactly. Anthony Lewis.
Anthony Lewis: This is a very, very simple point, but it's
just brought to mind an introductory phonology course. For the midterm, I'll
always give a problem that includes things like final devoicing and
syllable-related phenomena, and I'll give them some data from Spanish or
something where they have a phenomenon like that. And on a couple of occasions,
the students don't get the right answer in saying that this or that occurs in
syllable-final position, because they're just not trained to look or they don't
have that concept. And they give these correct answers: it occurs in front of
/b/, /t/, /s/, and they make this list of all these things. And I try to tell
them, "No, you're wrong. This is a syllable-final phenomenon and you didn't get
it." They say, "But, I'm absolutely right." This is really, it's over
simplistic but it's the very core of what they're saying.
Chuck Cairns: Do you mark them off?
Anthony Lewis: They're not trained to look at this
constituent whatsoever, so they're looking for explanations that could be just
as valid. So it's hard to tell them that they got the answer wrong.
Chuck Cairns: Bridget, did you want to say something?
Bridget Samuels: I guess so. I think the hard thing is: what
really do we mean when we say the syllable exists? Back to what Dasha was
saying, I mean, I don't mean to be existential here, like what do we mean by
exit? But seriously, to have a UG structure is one thing, and to say that
there's some kind of emergent thing is another thing, and it's really hard for
me to tell what kind of evidence would argue for one and not the other.
Shanti Ulfsbjoerninn: That was a really nice condensed way of
saying what I wanted to try to explain.
Donca Steriade: I think that the one thing that could argue
for one thing against another is that the conditions for emergence are very
likely to be very difference across languages. So you suggest that some ways in
which syllables might emerge on an individual basis, that presumably the
experience of the language learner is going to be different depending on
whether they are speaking Bella Coola, or Berber or Italian. So then, one would
Bridget Samuels: But what's the difference...
Donca Steriade: So then the kinds of things that are
emerging from that experience are going to be vastly different. But what I'm
struck by is that speakers who learn Tashlhiyt on the Dell and Elmedlaoui
analysis, the facts of the quantitative meter in Tashlhiyt are not emerging
from a sonorific wave theory of the syllable. They are emerging from something
very abstract, in which syllabic consonants are as good as vowels. They might
be a /p/ but they are as good as an /a/. So this is something that we should
bear in mind when we talk about emergence. Emergence is a great thing but it
has to be that we need to explain the emergence of the rhythm in Tashlhiyt,
Bridget Samuels: Sure. Whether my theory is right or wrong
is totally orthogonal...
Donca Steriade: No, no, no. I wasn't ____ ...
Bridget Samuels: No, it's fine, but I think it's very hard
to compare... There's two kinds of syllable structure here. There's one where
every syllable is invariant and there's the one where syllable structure is
different in every language and you have to learn based on experience what
syllable structure is possible. Determining between that kind of theory and an
emerging theory, and the two kinds of structural theories, this is the thing ....
Bill Idsardi: They'll differ in their developmental
sequence. Whether you can test that is another question. But the theory, it's
just again, standard science. All these problems boil down to: what do you do
when you have two extensionally equivalent theories? And that's not, there's no
solution to that. There's no solution to that in physics. There's no solution
to that anywhere.
Stephanie Shattuck-Hufnagel: There is a next step though.
There is a next step.
Bill Idsardi: Well then you have arguments about the
intentional concepts. So you have extensional equivalents, and then you argue,
"oh, my theory's simpler," or whatever it is.
Stephanie Shattuck-Hufnagel: Well no, you can work it out.
Bill Idsardi: Sometimes. There are cases where there are no
more phenomena to find. Sure, like you get this in math all the time, where not
only do you have extensional equivalence, but you have things where you can
prove that they're entirely equivalent.
Stephanie Shattuck-Hufnagel: We're not facing that problem,
in my opinion.
Bill Idsardi: I hope not, but you know, there are cases
where you're going to go and you're going to say, "Well, the appropriate data
is going to be this," so like developmental sequence data, and that I can't
Jason Shaw: Well, I'd like to add a new phenomenon then...
Chuck Cairns: Jason, Jason Shaw.
Jason Shaw: ... that could be accounted for, which is patterns
of temporal coordination. So we had one [?] topic, but not much discussion
about time. A lot of discussion about hierarchical structure. So another
phenomenon that might help us to distinguish between theories by an empirical
basis is potentially how prosodic structure unfolds in time. So I would add
temporal coordination of articulatory gestures to the set of... I mean, there's
very little data. We know some things. We know C-center effect in English, in
Georgian, not in Berber. I will tell you some more about Moroccan Arabic from
my poster if you care to stop by tomorrow. There are very little data in this
area but it's something that is an empirical domain that is not tapped and
potentially bears on this issue, so put it on the list.
Chuck Cairns: Will do.
Eric Raimy: I will add another phenomenon. So as far more
recent psycholinguistic evidence has found frequency effects that have priming
effects for syllables. The only way that I can understand that is if they're
memorized somehow, just because that's how I interpret the frequency and
priming effects on roots in Semitic templates and segments. So if we broaden
again our types of evidence that we're willing to accept, we have this
situation that either we can ignore the priming evidence and frequency evidence
for syllables, but then we have to kind of ignore that information for segments
and all the other things in that ____. So then it's forcing us to ask the
question of, if we wanted to get rid of the syllable, how could we get rid of
the syllable, but then still have those documented things?
Stephanie Shattuck-Hufnagel: Do you think it's absolutely
Eric Raimy: No,
and Joanna Cholin, who I believe has some involvement with that research
unfortunately walked out of here a few minutes ago, so...
Chuck Cairns: I think she just left. She didn't just walk
out – walk out. Right, right.
Eric Raimy: She looked like she had some place to be.
Chuck Cairns: Right, okay. Just checking.
Eric Raimy: No, as I said, I believe that these results are
just beginning to accrue, and so as with any other results, they need to be
replicated and looked at and reconsidered and treated with the proper care.
Stephanie Shattuck-Hufnagel: I'd be interested, for example,
if there's any way to test whether a foot frequency effect might exist in
Chuck Cairns: Did you say...
Stephanie Shattuck-Hufnagel: Foot.
Chuck Cairns: Oh, foot. What we're discussing might be the
topic of next year's conference, so... Harry.
Harry van der Hulst: Another thing on the list, or at least,
on my list that I want to correct an impression that I may have made yesterday
when Francois Dell's work came up, I said, "That can't be right," or something
like that. Of course what I meant was that if I'm right, or on the right track,
then my challenge is to account for these generalizations that they make in
that work in a way that is at least comparable in terms of explanatory value.
So I wasn't saying that these things couldn't be right or that these
generalizations aren't right. It was my challenge and my responsibility to show
that if I assume that there are these syllabic constituents or these particular
constraints, then I have to show that I can come up with the same... with an
account of these generalizations that is at least comparable. So that's what I
will put on my list, because I realize that a lot of what we say about
syllables is based on a fairly limited set of languages, either languages that
have hardly any structure at all, CV languages, or languages like Dutch and
other Indo-European languages that have these particular types of clusters. So
that's a big challenge, to look at these languages that are in other families
and that... again, if anybody who talks about syllables is right, then anybody
that says that you couldn't have an obstruent being the nucleus or the head of
the syllable, then I'd better deal with these generalizations.
Chuck Cairns: I think I saw another... yes.
Vsevolad Kapatinski: I'd like to add another phenomenon that
a theory of constituency or lack thereof could also explain. And the basic
difference between a theory that says that there is a constituent for the
syllable and a theory that says that there isn't is that a theory that says
that there is says that there is some kind of a separate representation for a
linguistic chunk, for the syllable, and the listeners are therefore
particularly likely to parse out that chunk as an acoustic segment, and also
that it's going to be easier for them to form an association between that chunk
and some other chunks, as opposed to between some random string of acoustic
signals and some other chunks.
Francois Dell: I have a hard time understanding this point,
I think, that Harry made ____ yesterday. Even if we never find any correlates
in word boundaries in a language, that wouldn't force us to abandon the notion
of the word. Really, it's something that, if we find some phonetic evidence,
some phonetic cues for syllable boundaries, so much the better. But if we
don't, I don't see what follows from that. The evidence for syllables still
Vsevolad Kapatinski: I wasn't talking about phonetic cues
for syllables. I was rather thinking about psychological imposition of
structure on the acoustic signal that listeners do in terms of parsing the
acoustic signal into chunks that they can store in their memories.
Bill Idsardi: There is, in fact, a tremendous amount of
evidence about that, but it's entirely equivocal. So if you look at morpheme
identification, the question is, are sub-syllabic morphemes like the English
plural more or less likely as constructs as opposed to full syllable ____
identifications of morphemes. And the evidence is that kids have no particular
problem with sub-syllabic morphemes. But that's psychological identification of
a unit, a pairing of a unit with another unit, and I don't see that there's any
huge bias to say that the morphemes are going to have to coincide with the
syllable boundaries. Yet there's something where we know absolutely that they
store them in memory. So, the evidence is entirely equivocal as to what that
sort of psycholinguistic technique which you want to use... is it going to
diagnose, is it going to be easier to pair things that are syllables with a
meaning in this case, or is it not? Well, the problem is there are all sorts of
other parses available, so it could parse as a segment so the question is, why
should the syllable have any privileged status with respect to any particular
Vsevolad Kapatinski: Right, it's true that there are other
units that could also be parsed out of the signal, so you could parse out the
morpheme or you could parse out the syllable, and when those parses conflict,
then you could go with either one. But if there is a situation where a
particular string of the acoustic signal doesn't correspond to any constituent,
then it seems to be very unlikely that a listener could eventually parse that out.
And in my poster yesterday, for instance, I had shown that in English
syllables, listeners find it much easier to learn associations within rimes and
affixes than bodies and affixes. They are more likely to parse out the rime
than the body.
Anthony Lewis: I think it's equivalent to saying that you
can have a syntactic constituent of five individual words as an important unit,
as a syntactic constituent, as opposed to say, two words from this type of
phrase and two from another, and at the same time still honor that you can have
the individual units that make up that syntactic constituent still the
meaningful units _____.
Ranjan Sen(?): I would like to say again that when we're
talking about the relationship between whatever constituent we want to argue
for or against and phonetics, we should always remember that phonetics isn't
just the acoustic signal. So we can look at the acoustic signal and see peaks
and troughs and vowel-ness, which correspond to the syllable, but then again,
we will not, when we look at the acoustic signal, readily find evidence for the
kind of abstract planning between consonant and vowel gestures, which we are
now getting good evidence for, which Jason just mentioned. So merely looking at
the acoustic signal is not exhausting the ways to investigate whether something
has a correlate in phonetics. We should be careful not to interconnect _____.
Shanti Ulfsbjoerninn: But if you presuppose an underlying
structure, let's say, like the strict CV as they branch down _____ or you just
have CVCVCVCV, then that whole phonetic planning is only like circumstantial
evidence, in the sense that you get languages in which that is... you get a nice
kind of relationship between, or a nice contrast between C and V. But then you
go _____ in which you just have your CVCVCV and it looks like you can stick
almost anything in those.
Ranjan Sen: I'm not talking about theories that presuppose...
this is very sophisticated argumentation for why we would say that gestures and
planning are this way and not that, which are now coming out of studies like
Jason Shaw: You can't _____ At some level of abstraction,
you might see the same type of priming relationship between an onset and a
nucleus in Berber, where both are consonants, and a /p/ and an /a/ in another
language. If you're willing to look at that level of abstraction, then it's at
least a hypothesis that's out there that there is some level at which you can
prime, kind of similarities in the priming relation that reflect some higher
Chuck Cairns: Can we maybe modularize the discussion for the
point of view of further in the discussion online and into the future? We've
been talking about syllable-related phenomena at a number of different levels,
and Paul started off the conversation by talking about the kinds of phenomena
that have traditionally been sort of within the core sets of phenomena that
phonology has tried to explain: phonotactics, linguistic phonetic sort of stuff
like aspiration, glottalization, typology, allomorph selection, and things sort
of like that, as opposed to more concrete things like phonetic planning. Is it
useful to distinguish, to make that kind of distinction, and to sort of
modularize the discussion so that it's not necessarily the case that the set of
arguments that you're going to use for one are going to be the same as the set
of arguments that you're going to use for the other? What do people think of
that kind of distinction?
Stephanie Shattuck-Hufagel: I think you need a third
category called "other."
Chuck Cairns. Yeah. Yes.
Unknown: I think that one point also is that the discussion
is getting difficult because we're talking about something we never defined
clearly. The only thing we can be sure about talking about syllables is that
everyone almost in this room would agree that CV is a syllable. The remaining,
it's totally unclear. Whether CVC is a syllable, CCV is a syllable, or two or
three, we don't know. So I think one starting point could be asking people to
challenge themselves in saying, if you claim that CCV is a syllable, then you
have to make the point clear that it is really a syllable. The same for CVC,
and then... otherwise we will be talking, talking, talking, nobody will be able
to challenge each analysis against each other.
Harry van der Hulst: No, but there is no agreement that even
CV is a syllable.
Chuck Cairns: Does anybody disagree that even CV is a
Unknown 3: It's just a sequence.
Unknown 4: It could be.
Chuck Cairns: Okay. But does anybody say, "no"?
Anthony Lewis: I would say no in light of San's paper, where
you have two segments being judged as one, so are you defining the C of your CV
as possibly consisting of two segments? I really liked his analysis as looking
at that onset consonant as a combination of _____ gestures, or something like
that, so I don't think there is a unanimous take on that.
Chuck Cairns: San?
San Duanmu: The onset is a mystery. In my analysis, the
metrical principle is true, and there is a trochaic foot only, so CVX is a free
C and VX is a troche and CV, it is hard to say. But the onset doesn't seem to
be required as part of the syllable. It just was something thrown out by CV
syllables, so it looks like it's hanging around, and I'm fine with it being
part of the syllable, but if you really think very carefully, I think I cannot
understand why you need that C there. But I want to make a point. I think we
probably are talking about issues that are too philosophical here, that will
actually hinder student research, because in a biology department, nobody talks
about what is life. So we actually, perhaps it would be more useful to think
about what are the things that are useful for us to do that will either
generate new data or it will simplify some previous theory. And then talk about
these afterwards. So, I don't want to _____ my own new research. I didn't
really want to assume there is a syllable, but I just go through the
distribution patterns and see how hard it has to be, and I end up seeing that
it's really CVX, and that is a surprise. In the end, I didn't claim that it
must be some fundamental unit. It could be derivable from a new structure, and
I haven't done the work on that side, so I can't claim, but I think there's
progress here. I didn't even try to define what the syllable is, but these
kinds of things would be very useful. I'm not bothered by what a syllable is,
but by not answering that question, you end up perhaps learning more to the
Eric Raimy: So what you're suggesting is massive amounts of
homework on the syllable for all our students.
San Duanmu: It's hard to say. You know, it's hard to say
ahead of time what to look at and I tries a lot of other things that didn't
work out, but I think progress can be made in different ways. You keep thinking
what the definition is and try to come up with a definition, but that's only
one way of... it's like philosophy. Then there is danger there.
Eric Raimy: Right, no, I was joking. I think this is the
same point that Stephanie made. It's that we appear to be in this conundrum
that we don't know exactly what to do, so we have to take it in steps. If we
don't know what to do, then I think we have to be very ecumenical, which is the
term that was used yesterday, to allow, just, we need to take steps. So I think
that's a great suggestion: that we need to just do more work so we can actually
work on the problem.
Ranjan Sen(?): Can I draw on an aspect of that last point
and something that Paul said previously? As a person who, I call myself a
phonologist, but also I work in the field, and so I have to describe the
language that I'm working with, people like me want tools and so a very good
thing to do, and maybe something to put on our list is, as we develop our
theories, don't only say what we've managed to describe, but say what we'd like
to know better that's out there, because if I'm going to describe the phonology
of a language, I don't have time to cover the fifty latest theories of the
syllable or the non-syllable, plus everything that's happened in the last fifty
years. But if I do know that there's a list of things which people want to know
about, I can say, "well, I've got some data that here that's relevant for
this." So it's useful, obviously not immediately for yourself to say, "I'd like
to know about this," but if you do say that, then there will be other people
like me around who will come to you with the data you may want. And what you
are doing will be much more useful.
Eric Raimy: It's 5:30.
Chuck Cairns: 5:30, good. Paul.
Paul Kiparsky: I thought I'd tell you a story. I'm quite
possibly the only person here who was present at the reinvention of the
syllable in 1966. And I do know that it wasn't there in the Jakobson feature
system and that Chomsky and Halle tried to write the Sound Pattern of English
without the syllable and then they discovered that they couldn't do it, so
finally in 1966, I think it was in the fall, they decided, "let's have the
feature" plus or minus syllabic. The wrong move, but in some sense it was
Chuck Cairns: ... the only thing they could do.
Paul Kiparsky: And it was immediately a great hit. Everybody
was saying there were all sorts of types of things you could do with it. I
remember in particular that there was a French visitor, Jean-Claude Milner, who
came and said, "ah, now I have what I need for the theory of French verse." [laughs in the audience]
Francois Dell: You gave him a German accent. [laughs in the audience]
Paul Kiparsky: Sorry about that. [laughs in the audience]
Chuck Cairns: So, we have the room until six. Do people want
to go further? We can do that. We have the restaurant reservation at six, so do
people want to go further now? I see people are taking leave-taking gestures,
so I guess we've come to a natural conclusion of the discussion. Thanks.
Total Visitors: 3127